Self-Driving & Driver-Assist Accident Claims

Two things are simultaneously true about a crash involving driver assistance, and they pull in opposite directions. The driver almost certainly remained legally responsible for the driving task. And the evidence that would show whether the system contributed sits with the manufacturer, not with either driver.

Quick answer: Nearly every consumer system marketed as self-driving is Level 2 under SAE J3016 — partial automation requiring continuous supervision, which keeps the driver legally responsible for the driving task. “The car was driving” is not a defence. But a product liability claim against the manufacturer can run alongside the ordinary negligence claim, on design defect, failure to warn, or marketing that overstated capability. The decisive evidence — engagement history, driver-monitoring data, takeover warnings, sensor and camera logs — is largely held by the manufacturer, and the software may have been updated since.

The Level Question Comes First

SAE J3016 defines six levels of driving automation, and the only one that matters at the outset is which one your vehicle was operating at:

• Levels 1–2 — driver support. The system assists with steering, speed or both. The human remains responsible for the driving task and must supervise continuously. Adaptive cruise control combined with lane centring is Level 2, and this is what nearly every consumer system on the road today is — including several marketed with names strongly implying more.
• Level 3 — conditional automation. The system performs the driving task within defined conditions, but the driver must remain available to resume when prompted. Availability in consumer vehicles is limited and geographically constrained.
• Levels 4–5 — high and full automation. No human takeover required within the system’s operational design domain. In practice today this means commercial robotaxi and shuttle operations rather than privately owned cars.

The reason this is the threshold question is simple: at Level 2, fault allocation still runs through the driver. The name of the feature, the marketing, and the driver’s belief about what the system could do change none of that as a matter of responsibility for the driving task.

What they can change is whether the manufacturer is also liable.

Two Claims, Not One

Against the driver — ordinary negligence, exactly as our guide to what to do after a car accident and the fault analysis in our guide to comparative negligence by state describe. A driver who was inattentive, took their hands off the wheel, or ignored takeover warnings was negligent whether or not a system was engaged. Where the driver was working at the time, the employer may be reachable through the doctrines in our guide to vicarious liability.

Against the manufacturer — product liability, on the three theories set out in our guide to defective product claims:

• Design defect — the system behaved unreasonably dangerously in a scenario it should have handled, or failed to detect something a reasonable system design would have detected. A safer feasible alternative design argument here often points at what competitors’ systems did, or at driver-monitoring capability the manufacturer had available and did not deploy.
• Manufacturing defect — a specific sensor, camera, radar unit or wiring assembly on this vehicle deviated from specification.
• Failure to warn — the system’s real operational limits were not adequately communicated, or the warnings were inconsistent with how the product was presented.
• Marketing and capability representations — a distinct strand of this litigation, resting on the principle that a manufacturer must account for foreseeable misuse, including misuse its own presentation of the product encouraged.

The two claims are not in tension. A driver can be primarily at fault while a manufacturer is also liable for a system that contributed, and apportionment sorts out the shares.

The Evidence Problem Is Different Here

In an ordinary collision, the evidence is at the scene and in the vehicles. Here, much of it is on someone else’s servers.

What the vehicle itself holds:

• The event data recorder, governed by 49 CFR Part 563 — speed, braking, throttle, restraint status in the seconds around impact. As our guide to truck accident black box data notes, the Driver Privacy Act of 2015 establishes that EDR data belongs to the vehicle’s owner or lessee, and retrieval generally requires consent or legal process.

What the manufacturer holds — and this is the bigger set:

• System engagement and disengagement history — whether the feature was on, when it turned off, and whether it disengaged before or because of the impact
• Takeover warnings issued and how long before the crash
• Driver-monitoring data — hands-on-wheel detection, eye tracking or cabin camera data where fitted
• Camera and sensor logs, including what the system detected and classified
• Telematics uploads, which may be the only surviving copy
• The exact software and firmware version in force at that moment

Two features make this fragile. First, retention is governed by the manufacturer’s own policies, which you cannot see. Second, over-the-air updates mean the software on the vehicle changes: the version that crashed may be gone, which complicates both reconstruction and the defect analysis. Establishing and preserving the version history is part of the case, not a technicality.

Federal Crash Reporting as Notice Evidence

NHTSA’s Standing General Order requires manufacturers and operators to report crashes involving automated driving systems and Level 2 driver-assist systems, and the reported incident data is published.

No individual claim is won on that data. What it can do is establish notice — that a manufacturer knew about a recurring failure pattern and continued without addressing it. That is the same evidentiary function that prior-incident history performs in a premises case, and it is what can move a design defect claim from arguable to strong, and support the conduct analysis in our guide to punitive damages.

Also worth checking: recall and technical service bulletin history for the specific system and model year, and any software recall, which for these vehicles may have been delivered as an update rather than a workshop visit.

Robotaxis and Driverless Operation

Where there is no driver, the claim changes shape entirely. There is no human driving task to allocate, so the defendants are:

• The company operating the service, on its own negligence in deployment, route selection, monitoring and response
• The manufacturer of the automated driving system
• A remote operator or remote assistance provider, where the vehicle was being supervised or assisted
• Software, mapping or sensor suppliers, as component manufacturers

The records that matter are the operator’s: the operational design domain the vehicle was permitted to work within and whether it was operating inside it, prior disengagements or incidents at that location, remote assistance logs, and the fleet’s own incident history.

On the regulatory position: federal motor vehicle safety standards were written around human-driven vehicles and do not comprehensively regulate automated driving systems, so there is no broad federal preemption defence of the kind that dominates our guide to train accident and grade crossing claims. Manufacturers argue regulatory compliance, but compliance is not the same as a preemption bar.

Practical Steps

1. Establish which system was engaged and at what level, from the vehicle’s own records rather than from anyone’s recollection.
2. Send a preservation demand to the manufacturer within days, naming engagement history, driver-monitoring data, takeover warnings, camera and sensor logs, telematics and the software version in force.
3. Preserve the vehicle itself, unrepaired, since sensors and their mounting are physical evidence.
4. Do not accept an over-the-air update to the vehicle while the claim is being investigated, if you can avoid it, and record the version currently installed.
5. Photograph the scene as normal, including road markings, lane lines, signage and lighting — Level 2 systems depend on exactly those inputs, so their condition is evidence.
6. Check NHTSA reporting, recall and bulletin history for that system and model year.
7. Do not concede “the car was driving” as either an excuse or an explanation — establish what the system actually did.
8. Get an attorney with product liability experience, not only motor vehicle experience, since the manufacturer claim is the half that needs it.

Sources & Further Reading

• SAE J3016 — Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles, the source of the Level 0–5 classification and of the allocation of the driving task at each level
• NHTSA Standing General Order on crash reporting for automated driving systems and Level 2 advanced driver assistance systems, and the published incident data collected under it
• 49 CFR Part 563 — event data recorder requirements; Driver Privacy Act of 2015 (part of the FAST Act) — establishing that EDR data belongs to the vehicle’s owner or lessee
• Restatement (Third) of Torts: Products Liability — design defect, manufacturing defect and failure to warn, including the treatment of foreseeable misuse
• NHTSA recall, investigation and technical service bulletin records for the specific system and model year, including software-delivered recalls
• State law on comparative fault apportionment between a negligent driver and a manufacturer, which determines how shared responsibility is divided
• See our guides to defective product claims for the manufacturer theories, truck accident black box data for how vehicle data is obtained and the ownership rules that apply, and vicarious liability for reaching an employer where the driver was working